What "AI-powered" should actually mean for a small shop
Most AI pitches to small businesses sell autonomy nobody asked for. Here is the model I would actually deploy — drafts, not decisions — worked through for a North Bay service business.
Notes on DevOps, incident response, and building sustainable engineering cultures. Short, opinionated, and written from direct experience — not the LinkedIn thought-leader version.
Most AI pitches to small businesses sell autonomy nobody asked for. Here is the model I would actually deploy — drafts, not decisions — worked through for a North Bay service business.
Moving self-hosted SSO from password plus TOTP to passkeys, and the real gotchas that show up when Windows Hello meets a Cloudflare Tunnel and Caddy stack.
I queued an offline-first warranty tracker as an overnight agent-fleet run and spent the next morning fixing what the agents got wrong.
Six hidden failures that blocked a watchOS target, and what solo iOS shippers should pre-provision before adding their first watch extension.
An expensive reviewer turns a cheap worker's mistakes into durable lessons, then injects them back into future prompts so the stateless model keeps getting better.
How I wired a self-hosted Signal bridge to a local AI assistant so the family can text it, and then gave it vision for photos.
Agents for throughput, a human as the reliability gate — how one operator ships iOS and Android apps to production.
Why hallucination rates differ wildly between models and why adjudication, not raw output, is the real product of a forensic AI audit.
S/MIME end-to-end with board-owned keys — structural privacy, not policy privacy.
Turn incidents into durable improvements without witch hunts or wheel-spinning.
How I help strong ICs become calm, trusted incident leaders.
Clarified scope, severity levels, and paging rules.
─────────────────────────────────────────────────────
Found this useful? Send me a note.